shows characteristics of the 3,075 participants in Health ABC at enrollment (1997–1998). The mean age of the cohort was 73.6 years. Race was based on self-report and 41.7% of the cohort was black. Although participants were well functioning, comorbid conditions, such as diabetes (15.3%), coronary heart disease (20.8%), and reduced lung function (29.3%), were common.
Three hundred and four cases of CAP requiring hospitalization occurred over the 12-year ascertainment period from 1992–1993 to 2005. Of these, 106 cases occurred in 106 subjects during the 5 years before enrollment in Health ABC, and 237 cases occurred in 198 subjects after enrollment. Thus, 272 subjects (8.8%) were hospitalized at least once for CAP during the 12-year ascertainment period and 29 participants had more than one episode of CAP hospitalization. The rate of CAP after enrollment in Health ABC was 10.9 cases per 1,000 person-years. compares clinical characteristics and inflammatory marker concentrations for participants with and without at least one hospitalization for CAP. Participants with CAP were more likely to have comorbid conditions, such as congestive heart failure, coronary artery disease, and diabetes, elevated serum creatinine concentrations, and reduced lung function; in addition, they were more likely to be or to have been smokers, and to have higher circulating concentrations of TNF and IL-6. We measured PAI-1 concentrations in 3,010 (98%) participants at enrollment, when subjects reported good health and absence of infection. The median PAI-1 concentration was 21 ng/ml. Circulating PAI-1 concentrations were higher among subjects hospitalized for CAP compared with those never hospitalized for CAP (24 vs. 21 ng/ml, respectively; P = 0.02).
Of the 272 participants with hospitalization for CAP, 162 (60%) were white and 110 (40%) were black. compares characteristics of participants with and without an episode of CAP stratified by self-reported race. In this stratified analysis, the incidence of CAP was again higher in participants with comorbid conditions and a history of smoking, but results did not reach statistical significance with some of these comorbid conditions. PAI-1 concentrations were higher among whites hospitalized with CAP (25 vs. 21 ng/ml for hospitalized and nonhospitalized subjects, respectively; P = 0.005), but differences were smaller and not statistically significant among blacks who did and did not require hospitalization for CAP (22 vs. 21 ng/ml, respectively; P = 0.7).
CHARACTERISTICS OF COHORT STRATIFIED BY SELF-REPORTED RACE
PAI-1 SNPs and Haplotypes
All SNPs were in Hardy-Weinberg equilibrium, except the PAI2852 in whites and PAI4588 and PAI12219 in blacks (). Compared with whites, blacks are genetically more heterogeneous. Therefore, we analyzed markers for European ancestry among 1,278 (97%) blacks. Because racial admixture could account for Hardy-Weinberg disequilibrium, we estimated Hardy-Weinberg equilibrium among blacks after stratifying for European ancestry markers. All SNPs were in Hardy-Weinberg equilibrium among blacks with less than 10% of markers of European ancestry. Hardy-Weinberg disequilibrium was present for the PAI12219 SNP and those with greater than 25% of markers of European ancestry and for both PAI4588 and PAI12219 SNPs and those with greater than 50% of markers of European ancestry.
TEST FOR HARDY-WEINBERG EQUILIBRIUM (EXACT P VALUES) FOR PAI-1 GENOTYPES STRATIFIED BY SELF-REPORTED RACE AND BY PERCENTAGE OF EUROPEAN ANCESTRY FOR BLACKS
Measures of linkage disequilibrium between the PAI-1 SNPs are shown in Table E2A. We constructed nine haplotypes (frequency > 1%) for the entire cohort, and the frequency of these haplotypes varied between blacks and whites (). Although haplotypes A, D, and E accounted for 69.5% of the haplotypes among whites, haplotypes A, B, and C accounted for 76.8% of haplotypes among blacks. Haplotype B was present among 34.1% of black participants, but only 2.2% of whites. In contrast, haplotypes D and E were present in 16.3 and 10.7% of white participants but in only 4.1 and 1.9% of blacks, respectively.
FREQUENCY OF HAPLOTYPES BASED ON SEVEN SINGLE NUCLEOTIDE POLYMORPHISMS*
PAI-1 SNPs and CAP Susceptibility
Among white participants, variants at the PAI4G,5G, PAI2846, and PAI7343 sites had higher risk of CAP (P = 0.018, 0.021, and 0.021, respectively; ). These associations remained significant after adjusting for false discovery. A codominant effect was seen for the 4G allele at the PAI4G,5G site (frequency of CAP = 9.9% vs. 5.3% for the combined 4G/4G and 4G/5G genotypes compared with 5G/5G genotypes, P = 0.005), and those with at least one 4G allele had a 1.98-fold odds of CAP hospitalization (95% confidence interval [CI], 1.2–3.2; P = 0.006). The odds of CAP for white participants with the 4G/4G and 4G/5G genotypes remained unchanged (OR, 2.1; 95% CI, 1.3–3.6) in the multivariable analysis. Similarly, a codominant effect was seen for the AG and GG genotypes at the PAI7343 site (frequency of CAP = 6.5% vs. 10.5%, for the AG and GG genotypes compared with AA genotypes, P = 0.006). Subjects with at least one G allele at the PAI7343 site had a lower risk of CAP in univariate and multivariable analyses (OR, 0.6; 95% CI, 0.4–0.9; P = 0.006; and OR, 0.5; 95% CI, 0.4–0.8; P = 0.001; in univariate and multivariable analyses). A log-additive effect was seen for the genotypes at the PAI2846 site () with the highest frequency of CAP for subjects with the GG genotype (9.9%), an intermediate frequency for those with the GA genotype (5.6%), and the lowest frequency for the AA genotype (0%) (P = 0.01), but the overall frequency of subjects with the AA genotype was very low (1.4%).
ASSOCIATION BETWEEN GENOTYPE AND PHENOTYPE (PAI-1 AND COMMUNITY-ACQUIRED PNEUMONIA)
In contrast to white participants, only genotypes at the PAI12750 site were associated with CAP susceptibility. Subjects with the AA genotype were associated with the highest frequency of CAP (50%) compared with AG and GG genotype (6.1 and 9.5%, P = 0.01), but only 0.3% of black participants had the AA genotype.
PAI-1 SNPs and Circulating PAI-1 Concentrations
In general, genotypes associated with higher risk of CAP among whites had higher circulating PAI-1 concentrations (). For example, those with the 4G/4G and 4G/5G genotypes had higher PAI-1 concentrations compared with the 5G/5G genotypes (23 vs. 20 ng/ml, P < 0.0001). Similarly, subjects with the AA genotypes at the PAI7343 site had higher circulating concentrations compared with the AG and GG genotypes (22 vs. 20 ng/ml, P = 0.02).
We stratified our analyses by site to assess for population stratification (). Again, genotypes associated with higher risk of CAP at Memphis and Pittsburgh sites were associated with higher circulating PAI-1 concentrations. Although the direction of association was similar at both sites, the magnitude varied. For instance, genotypes associated with higher circulating PAI-1 concentrations at the PAI4G,5G and PAI12219 sites were associated with higher risk of CAP at both sites, but the associations were statistically significant only at the Memphis site. Similarly, genotypes associated with higher circulating concentrations at the PAI2846 and PAI7343 sites were associated with higher risk of CAP at both sites, but the associations were statistically significant at the Pittsburgh site only.
ASSOCIATION BETWEEN GENOTYPE AND PHENOTYPE (PAI-1 AND COMMUNITY-ACQUIRED PNEUMONIA) IN WHITES STRATIFIED BY SITE
PAI-1 SNPs and Cytokine Expression in Ex Vivo Whole Blood Stimulation
We assessed the functional significance of PAI4G,5G, PAI2846, and PAI7343 genotypes using ex vivo whole blood stimulation with LPS and PGN in 23 healthy white volunteers because these genotypes were associated with CAP susceptibility among white participants (). Only the PAI4G,5G polymorphism was associated with increased PAI-1 expression. Stimulation of the whole blood with LPS and PGN resulted in a 3.3- and 1.9-fold higher PAI-1 expression when 4G/4G and 4G/5G genotypes were compared with 4G/4G genotypes at the PAI 4G,5G site (1 vs. 0.3 ng/ml, P = 0.043, and 2.4 vs. 1.3 ng/ml, P = 0.034, for LPS and PGN, respectively) (). However, no differences were seen in TNF, IL-6, and IL-10 expression for the PAI4G,5G genotypes. Stimulation with PGN resulted in higher TNF expression for the AA and AG genotypes when compared with the GG genotypes at PAI2846 site (P = 0.001), but no differences were seen in TNF expression after LPS stimulation and in PAI-1, IL-6, and IL-10 expression after both LPS and PGN stimulation (). No associations were seen between PAI7343 genotypes and either cytokine expression after LPS and PGN stimulation ().
Figure 2. Changes in plasminogen activator inhibitor (PAI)-1, tumor necrosis factor (TNF), IL-6, and IL-10 concentrations after ex vivo whole blood stimulation with lipopolysaccharide (LPS) and peptidoglycan (PGN) for PAI4G,5G (A), PAI2846 (B), and PAI7343 (C) (more ...)
PAI-1 Haplotypes and Phenotypes
Although we genotyped all seven SNPs in all subjects, the PAI4G,5G, PAI2846, PAI4588, and PAI 7343 SNPs could tag the five common haplotypes (A, C, D, E, F) in whites. Of the 53% of white participants with a 4G allele, 46% had the 4G/G/C/A haplotype, which was associated with higher odds of CAP (P = 0.004) (). Of the remaining 47% of white participants with the 5G allele, only the 5G/A/C/G haplotype was associated with a lower frequency of CAP and this haplotype was only observed among 10.7% of white participants. Higher probability of the 5G/A/C/G haplotype was associated with lower risk of CAP (P = 0.004). However, higher probability of the 4G/G/C/A haplotype was not associated with higher risk of CAP (P = 0.06). A combination of three SNPs could tag the four common haplotypes (A, B, C, H) with a frequency greater than 5% in blacks (PAI4G,5G/PAI2846/PAI12750) (), but neither haplotype was associated with CAP susceptibility (P = 0.33).
ASSOCIATION BETWEEN COMMUNITY-ACQUIRED PNEUMONIA REQUIRING HOSPITALIZATION AND HAPLOTYPES BASED ON FOUR POLYMORPHISMS AMONG WHITE AND THREE POLYMORPHISMS AMONG BLACK PARTICIPANTS
Magnitude of Association between SNPs and Haplotypes with Phenotypes
In the Health ABC cohort, the PAI4G,5G alleles described only 1 and 0.3% of variation in the circulating PAI-1 concentrations measured in the healthy state in whites and blacks. Furthermore, the 4G/G/C/A and 5G/A/C/G haplotypes, associated with CAP susceptibility in whites, could explain only 1 and 0.1% of the variation in circulating PAI-1 concentrations measured in the healthy state. However, PAI4G,5G alleles described 20 and 22% of variation in PAI-1 expression after stimulation with LPS and PGN in healthy white volunteers. Finally, 43% of CAP cases among white subjects in the Health ABC cohort could be attributed to the 4G,5G alleles.
We conducted sensitivity analyses to assess associations between the PAI4G,5G SNP and the haplotypes with CAP among whites after excluding CAP events before enrollment in the Health ABC study. One hundred and ninety-eight participants (6.4%) had at least one hospitalization for CAP after enrollment in the Health ABC study: 126 were whites and 72 were blacks. Again, a codominant effect for the presence of a 4G allele at the PAI4G,5G site was seen in white participants. The OR for the combined 4G/4G and 4G/5G genotypes versus the 5G/5G genotype was similar to that when including all known cases in the analysis. Results did not reach statistical significance in univariate analysis (relative risk [RR], 1.6; 95% CI, 0.97–2.7; P = 0.069) but were significant in multivariable analysis (RR, 1.8; 95% CI, 1.03–3.2; P = 0.04). Haplotype analysis showed that the 4G/G/C/A haplotype was associated with lower odds of CAP susceptibility (P = 0.007), whereas the 5G/A/C/G haplotype approached, but did not reach, statistical significance (P = 0.06) (Table E3A).